Committed Consumption

Committed packages can be added to yearly subscriptions and run for the same duration as the license. They provide a cost-effective choice for steady, predictable consumption. If your needs exceed the package during the year, you can add Credits at any time.

Elastic Storage

Overview

Read more about Elastic Storage and how to calculate your yearly needs here: https://help.dectris.cloud/en/article/elastic-storage

Archiving

Overview

Archiving, also known as cold storage, holds data you want to keep but not actively process. Archived data is visible but must be restored to Elastic Storage before processing. This can be done in the DECTRIS CLOUD web app with one click. Retrieval typically takes up to 12 hours.

When to use Archiving and when Elastic Storage

The advantage of Archiving is the lower price per unit compared to Elastic Storage but comes with the limitation that data in Archiving cannot be processed. Thus, you want to keep the data in Elastic Storage as long as it is actively used to be processed and analyzed but move it as soon as possible to Archiving when not actively used. Also, to manage your Elastic Storage capacity it makes sense to move data fast to Archiving and reload it back to Elastic Storage when actually needed.

CPU Compute

Overview

1 CPUh corresponds to using 1 CPU for 1 hour. Using, for example, a machine with 64 CPUs for one hour would correspond to 64 CPUh.

How much CPUh do I need?

Assuming you spend 20h per week with data analysis in DECTRIS CLOUD mainly using machines with 8 CPUs and eventually one with 64 CPUs in parallel:

20h * 8 CPUs + 2h * 64 CPUs * 52 weeks ≅ 15 kCPUh/year

Or assuming you spend 25% of your time doing data analysis, you look at roughly 500 CPUh per month or 7500 kCPUh per year.

GPU Compute

Overview

1 GPUh corresponds to 1h of GPU use of NVIDIA A10 or L4. Higher performing GPUs might come at different rates.

How do I know if I need GPUs?

Check the requirements of the software you use. If a specific GPU type is listed (e.g., NVIDIA A10), you will likely need GPU compute time.